Cointegration
A statistical relationship where two time series share a long-run equilibrium — even if they wander apart short-term, they tend to revert to a stable relationship.
Explained Simply
Two stocks are cointegrated if their price ratio or spread is stationary (mean-reverting) even though each individual stock's price is non-stationary (can trend). This is the mathematical foundation of pairs trading. Cointegration is tested using the Engle-Granger or Johansen test. A lower p-value (below 0.05) means stronger cointegration. The concept is different from correlation: two stocks can be correlated without being cointegrated.
How Tradewink Uses Cointegration
Our PairsTrader module tests cointegration across 1,000+ potential stock pairs using the Engle-Granger method. Pairs with p-value <0.05 and favorable half-life (5-30 days) qualify for pairs trade signals. The AI also monitors cointegration stability over rolling windows — if cointegration breaks down, the pair is removed from the active signal universe.
See Cointegration in action
Tradewink uses cointegration as part of its AI trading signal pipeline. Start getting signals that use this concept to find real opportunities.